On knowledge-transfer characterization in dynamic attributed networks

How do social aspects influence knowledge transfer in dynamic attributed networks? We address this issue by characterizing the behavior of the actors and their dynamic interactions based on the strategic positioning in a social structure. For this, we propose a method to characterize the behavior of nodes and their dynamic relationships based on temporal node attributes that capture how knowledge is transferred across a network. In order to assess our method, we apply it to unveil the differences of social relationships in distinct academic social networks and Q&A communities. We also validate our social definitions considering the importance of the nodes and edges in a social structure by means of network properties, as well as investigate the robustness of our method by stressing it for dealing with the time existence of the nodes in a network and the randomness of attributes associated with them. Moreover, we propose an unsupervised method to measure the academic importance of researchers based on our knowledge-transfer model, which outperforms traditional network metrics and other social-based approaches.